NKG2D and its ligands are implicated in autoimmune diabetes in both humans and mice (Nikitina-Zake et al., 2004
; Ogasawara et al., 2004
). However, it was not clear how NKG2D and NKG2D ligands could contribute to diabetes development. Our current studies offer one possible mechanism. We demonstrate that NKG2D engagement on CTLs with its ligands expressed on islet cells can lead to the recruitment of the CTL to pancreatic islets. Through the secretion of chemokine, this can then lead to the recruitment of additional T cells that do not express NKG2D. This suggests that at least one function of NKG2D in diabetes may be to mediate the recruitment of T cells to pancreatic islets.
The mechanisms by which CTLs are recruited to inflammatory sites are only beginning to be characterized. Although once thought to have unrestricted access to all tissues, it is now becoming clear that CTLs require specific signals to enter certain organs. Signals described to date include specific chemokine expression in tissues following CD4+
T cell recruitment (Masopust et al., 2004
) and antigen-recognition (Wang et al., 2010
). Here we demonstrate a mechanism by which CTLs can be recruited to NKG2D ligand-expressing pancreatic islets. This recruitment was dependent on NKG2D expression by the CTLs and independent of cognate antigen recognition.
Although it was first realized that NKG2D was expressed on CD8+
T cells twelve years ago (Bauer et al., 1999
), the function of NKG2D on CD8+
T cells is still being delineated. NKG2D was originally described as a costimulatory receptor for CTL (Groh et al., 2001
), with no function in the absence of coincident TCR signaling. Since this time, NKG2D signaling on CTLs has been demonstrated to have some function in the absence of antigen recognition by the TCR under the correct conditions (Meresse et al., 2004
; Verneris et al., 2004
). Our early studies demonstrated that NKG2D engagement could induce immunological synapse formation in CTLs independent of antigen (Markiewicz et al., 2005
). However, the significance of this finding was unclear, as NKG2D-mediated synapse formation did not result in any measurable CTL effector function. Our present data suggest this may lead to the accumulation of CTLs in sites of NKG2D ligand expression in the absence of antigen recognition by the TCR.
Unexpectedly, recruitment of CTLs to islets was accompanied by a much larger recruitment of additional lymphocytes. Chemokine analysis showed that following T cell adoptive transfer, CCL5, a potent T cell chemoattractant, was induced in RAE1-expressing islets. Given that the CTLs themselves secreted a significant amount of CCL5, this was likely what was responsible for the recruitment of additional lymphocytes. We cannot conclude from our studies whether the additional lymphocytes recruited following NKG2D-mediated CTL recruitment were specific for islet antigens or not. Nonetheless, RAE1 expression by islets led to recruitment of cells triggered by an antigen-independent CTL response.
Although insulitis was not seen in younger RIP-RAE1ε mice, a mild insulitis did develop as the mice aged. The absence of insulitis in young mice suggests that despite constitutive expression of NKG2D by NK cells, NK cells are not able to traffic into RAE1-expressing islets by themselves and/or require additional factors to kill islet cells. Similar to our findings, Strid et al. (Strid et al., 2008
) demonstrated that acute upregulation of RAE1-β in the epidermis induced immune infiltration. However, in contrast to the slower infiltration of immune cells we observed in the pancreas of the RIP-RAE1ε mice, the acute expression of RAE1-β in skin rapidly induced immune infiltration of CD4−
αβ unconventional T cells, a cell type that we did not observe in the pancreas. It is not clear why the character of the infiltrate in response to RAE1 expression is so different between these two models. It could be a differential response in skin versus pancreas, to acute versus chronic RAE1 expression, or to different RAE1 isoforms. Additionally, as the receptor responsive to RAE1-β in the epidermis was not determined, the response seen in the skin may have involved a receptor in addition to NKG2D.
Because NKG2D is only expressed on CTLs, and not on naïve CD8+ T cells, we suspected that the delayed development of insulitis in our transgenic mice was due to low numbers of CTLs present in our mice, which are maintained in specific pathogen free conditions, and the absence of inflammation to promote cell entry. We tested these hypotheses by treating mice with a small dose of STZ to damage and inflame the islets. While this dose of STZ resulted in little to no cell infiltration in wild-type mice, this was able to induce inflammation in RAE1-expressing mice. To increase the number of CTLs, we infected mice with influenza virus. Infection resulted in increased numbers of infiltrating lymphocytes in the non-transgenic islet, suggesting that viral infection may have multiple effects including increasing CTL numbers that may allow for immune cell infiltration into peripheral organs including the pancreas. The expression of RAE1ε resulted in a moderately enhanced number of infiltrating lymphocytes when the mice were infected with influenza. Combining STZ and infection with influenza resulted in a synergistic effect with a much greater lymphocyte infiltration detected in RAE1ε-expressing animals compared to the single transgenic controls. This suggests that in the pathogenesis of diabetes, a virus that could infect the islets and induce NKG2D ligand expression might be able to both stimulate islet inflammation as well as induce CTL infiltration.
Analysis of the pancreatic cellular infiltrate in older RIP-RAE1ε mice demonstrated that CTLs constituted only a minority of the cells. Our results from the adoptive transfer studies allow for the possibility that it may be these cells that were originally recruited due to their expression of NKG2D and that they then recruited other cells via chemokines. This would suggest that the development of insulitis in our older mice may be determined not only by the infectious history of the mouse but also by an inflammatory trigger in the islets. However, our current data does not allow us to definitively determine the cell-type(s) driving this spontaneous recruitment of lymphocytes. In addition, while the influx of lymphocytes was able to generate appreciable inflammation, it, however, did not result in the development of diabetes. This suggests that additional factors, including the participation of antigen specific T cells capable of killing islet cells, may be required for the development of disease.
In contrast to our findings, two recent reports (Lennon et al., 2009
; Wang et al.) demonstrated that recruitment of T cells to islets was a cell-autonomous event that required islet-antigen specificity. The first (Lennon et al., 2009
) used an elegant system of retrogenic mice that expressed both islet-specific and non-specific CD4+
T cells. In this system there was no recruitment of T cells specific for irrelevant antigen to islets even when there was significant recruitment of diabetogenic T cells. One caveat to these studies is that the non-specific T cells were naïve. T cell activation alters receptor expression, including integrins and chemokine receptors, allowing for T cell trafficking to tissues (Denucci et al., 2009
). It may be that only activated T cells, or subsets of these cells, are able to be recruited in an antigen-independent fashion due to differential receptor expression. Our data suggest that NKG2D may be one of these receptors. In the mouse NKG2D is only expressed on activated CD8+
T cells (Raulet, 2003
), a cell type not investigated in the retrogenic mouse study.
The authors of the second study (Wang et al., 2010
) generated a gene targeted NOD mouse with a mutated islet-specific glucose-6-phosphatase catalytic subunit-related protein (IGRP) so that diabetogenic CD8+
T cells normally responsive to an epitope derived from IGRP were no longer responsive. These mice developed diabetes normally, but no IGRP-specific CD8+
T cells were found in inflamed islets. These authors did test the possibility that activation was required for recruitment of bystander T cells by adoptively transferring IGRP-specific CD8+
T cells activated in vitro for 3 days into the mice. These activated CD8+
T cells were also not recruited to the islets early after transfer. However, NKG2D expression is not upregulated on T cells early in culture (Markiewicz et al., 2005
). Consistent with our data, small numbers of these cells were recruited to islets by 5 days after transfer, at a time when they presumably expressed higher levels of NKG2D.
A major unanswered question is why NKG2D ligands would be expressed in pancreatic islets. Are there circumstances that induce expression of NKG2D ligands in islets normally, or does pancreatic expression of these proteins only occur in diabetic-prone individuals due to some dysregulation? It may be that NKG2D ligands are normally expressed during times of pancreatic stress, but diabetes only ensues when there are additional defects in immune tolerance resulting in the presence of cells capable of killing β-cells. One such stress may be the perinatal wave of β-cell death that is required for proper tissue remodeling of the pancreas and a proposed precipitating event for the development of diabetes (Finegood et al., 1995
; Kassem et al., 2000
; Petrik et al., 1998
; Scaglia et al., 1997
; Trudeau et al., 2000
; Turley et al., 2003
). Another pancreatic stress proposed to play a role in autoimmune diabetes development, viral infection (Hober and Sauter, 2010
), could also induce expression of NKG2D ligands. Future studies will be required to determine whether these or other stimuli induce expression of NKG2D ligands in normal or diabetic-prone individuals.
In summary, using transgenic mice with expression of the NKG2D ligand RAEε in pancreatic islets, we demonstrate that expression of NKG2D ligands in the pancreastic islets can induce the recruitment of CTLs to islets. This recruitment was independent of islet antigen-specificity, but dependent on NKG2D engagement on the CTLs. Further, once CTLs are recruited to islets, a large number of additional lymphocytes can be recruited via chemokine secretion. Given that aberrant NKG2D ligand expression has been linked to diabetes development (Ogasawara et al., 2004
), our results suggest that at least one role for NKG2D in the development of diabetes may be to mediate the recruitment of T cells to such pancreatic islets.